Solution Architecture

An AI medical-imaging intelligent-diagnosis robot system integrates a machine-learning library and a medical image-processing library into a deployment-free, integrated system. It includes most current medical image-processing and classification algorithms, fully ensuring practicality and advancement.

It currently includes two main diagnostic functions: disease recognition and similar-case retrieval. Disease recognition uses traditional pattern recognition and the latest deep learning to analyze and learn from input images and build diagnostic models. Models can evolve, continuously optimizing during diagnosis so they more closely reflect the real world and achieve high-precision diagnosis.

Lesion recognition and evaluation can identify and display key information such as lesion location and size and, following the internationally known RadLex disease-grade and disease-type semantic dictionary, automatically output diagnostic reports in a standardized format. Similar-case retrieval can find the most similar patient cases by features, providing reliable reference for the target patient’s diagnosis and treatment. In experimental tests on 3,820 cases, Sugon’s intelligent reading system reached a 90% recognition rate for 11 lung-CT diseases.